Date:         Thu, 19 Sep 2002 07:02:27 -0700
Reply-To:     Jeff Graves <jag@ST-ANDREWS.AC.UK>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jeff Graves <jag@ST-ANDREWS.AC.UK>
Organization: http://groups.google.com/
Subject:      cross over designs in mixed models
Content-Type: text/plain; charset=ISO-8859-1

Hello
We are new SAS users and are having a few problems with some of our
statistical analyses. The data are female birds laying a clutch of
eggs with attractive and unattractive mates. We are looking at the
mass of the eggs laid by females in both treatments. The design is a
cross-over one such that the females were given both types of mate,
(with half the females mated to the attractive male first, and half to
the unattractive male first). We have used a mixed GLM with egg mass
turned into a repeated measure for each female and using
autoregressive covariance (egg mass generally increases with laying
order). Clutch sizes vary and so sample sizes are unequal.
We want to analyse the data in two ways, firstly looking at just one
round of egg laying to look at the differences between the two
treatments, and then to look at the cross-over design in a paired
analysis with females that laid in both treatments.
We are confused as to whether or not to include a random statement
(for female) and what difference this makes. Secondly, after the model
statement, we included the line that is: repeated / type = ar(1)
sub=female(attractiveness) r rcorr;. This line appeared in the worked
example in the manual, but we are having trouble finding out what it
actually does.
Any help would be greatly appreciated
Alison Rutstein and Dr. Jeff Graves, University of St. Andrews